GLOBAL BIOGEOCHEMICAL CYCLES, VOL. 12, NO. 3, PAGES 479-499, SEPTEMBER 1998 Meridional transport of dissolvedinorganic carbon in the South Atlantic Ocean J. Holfort,1'2K. M. Johnson, 1B. Schneider, • G. Siedler, 4andD. W. R. Wallace1,2 Abstract. The meridionaloceanictransports of dissolved inorganiccarbonandoxygen were calculatedusingsix transoceanic sectionsoccupiedin the SouthAtlanticbetween 11øSand30øS. The totaldissolvedinorganiccarbon(TCO2)datawere interpolatedonto conductivity-temperature-depth datato obtaina high-resolutiondata set, and Ekman, depth-dependent anddepth-independent components of the transportwere estimated. Uncertainties in the depth-independent velocitydistribution werereducedusingan inversemodel. The inorganiccarbontransportbetween11øS and30øS was southward, decreased slightlytowardthesouth,andwas-2150 _+200 kmols4 (-0.81 _+0.08 Gt C yr4) at 20øS.Thisestimate includes thecontribution of netmasstransport requiredto balancethe salttransportthroughBeringStrait. Anthropogenic CO2concentrations were estimatedfor the sections.The meridionaltransportof anthropogenic CO2wasnorth- ward,increased towardthenorth,andwas 430 kmols-• (0.16 Gt C yr4) at 20øS.The calculations implynetsouthward inorganic carbontransport of 2580kmols4 (1 Gt C yr-•) duringpreindustrialtimes.The slightcontemporary convergence of inorganiccarbon between10øSand 30øSis balancedby storageof anthropogenic CO2and a sea-to-airflux implyinglittle local divergenceof the organiccarbontransport.During the preindustrial era, therewas significantregionalconvergence of bothinorganiccarbonandoxygen, consistent with a sea-to-airgasflux drivenby warming.The northwardtransportof anthropogenic CO2carriedby the meridionaloverturningcirculationrepresentsan importantsourcefor anthropogenic CO2currentlybeingstoredwithin the North Atlantic Ocean. 1. however,to be usefulin analyzingthe globalcarboncycle, suchan analysisrequiresan assessment of thecorresponding Introduction As high-qualitydataon the globaldistribution of CO2 in the atmospherebecameavailable, it becamepossibleto combinethesedatawithatmospheric transportmodelsto infer the transportof CO2 from onepart of the globeto another [e.g., Keelinget al., 1989; Tanset al., 1990;Entinget al., 1995]. This approachpromisesto provideinsightinto the regionaldistribution of sources andsinksof atmospheric CO2; transportof CO2 by the oceancirculation[Broeckerand Peng, 1992]. The oceanic transportof chemicalpropertiescan be approached usingmethodssimilarto thosedeveloped for the calculationof heatandwatertransports [e.g., Wunsch,1978; Hall andBryden,1982;Roemmich and Wunsch,1985].This approachhas been appliedto the transportof nutrients [Rintouland Wunsch,1991; Wunschet al., 1983; Robbins carbon[Breweret al., 1989; •Department of AppliedScience,Brookhaven NationalLaboratory, andBryden,1994]andinorganic Martel and Wunsch, 1993; Robbins, 1994; Holfort et al., 1994; $toll et al., 1996]. The latter made use of measurements of the total dissolvedinorganic carbon content of 3Sektion Meereschemie,Institut fOr Ostseeforschung,Rostock- seawater(TCO2). Warnemtlnde,Germany. Someof theseanalysesof inorganiccarbontransports have 4InstitutfOr Meereskundean der UniversitatKiel, Kiel Germany. been limited by a lack of high-quality,denselysampled Upton,New York. 2 Now at Institut fOr Meereskunde an der Universitat Kiel, Germany. Kiel, sections with TCO2 measurements, aswell asby incomplete specification of thevariouscomponents of thetransport which can contributeto the meridionaltransportof CO2. Densely sampled,high-qualityTCO2 dataare, however,now being collected in conjunctionwith World Ocean Circulation Experiment(WOCE) cruises,by theJointGlobalOceanFlux Study(JGOFS)globalsurveyof CO2in theoceans[Sabineet Copyright1998 by theAmericanGeophysical Union. Papernumber98GB01533. 0886-6236/98/98GB-01533 $12.00 479 480 HOLFORT ET AL.: SOUTH ATLANTIC CO2 TRANSPORT al., 1997]. An exampleof the use of transportestimates operatormultiparameter metabolicanalyzersystems [Johnson calculated usingmodernsection datawasrecentlyreported by and Wallace, 1992; Johnsonet a/.,1993]. On the WOCE $toll et al. [1996]. cruises,theTCO2 analyses werecross-checked for accuracy In thispaper, we use a setof thesenew WOCE sections, againstCertified ReferenceMaterials provided by A. withcorresponding TCO2data,asthebasisforcalculating the Dickson of Scripps Institutionof Oceanography.These north-south transport of CO2withintheSouthAtlanticOcean analysessuggested thatthe WOCE dataare accurateto better (we alsoreportthetransportof dissolved oxygen).We focus than 2 3tmolkg-•. Additionalsections were takenfrom a greatdealof anentionontheuncertainties whichinevitably historical data,particularly thehigh-quality pre-WOCETCO2 arisein suchcalculations. The calculations we presentare datacollectedduringthe SouthAtlanticVentilationExperiappropriatefor the early 1990swhen the observations were ment (SAVE) (T. Takahashiand D. Chipman,personal made.However,the concentration of CO2 in the oceansis communication,1991). We have comparedthe data from increasingwith time as a resultof the uptakeof excessor SAVE Legs2 and3 in theWesternBasinwith thetheWOCE anthropogenic CO2fromtheatmosphere, andanthropogenic TCO2 data from the sameregionusinga multivariateapCO2 concentrations and oceancurrentsare both spafially proachand foundno significantoffsetsbetweenthe datasets variable. We therefore also estimated the concentration [Wallaceet al., 1996]. Table 1 presentsa summaryof the distribution of anthropogenic CO2 in the sections andcalcu- sectionsused,and Figure 1 showstheir geographical localatedits transport.On the assumption thattheoceancircula- tions. tion andthe ocean'snaturalcarboncyclehavenot changed In orderto calculatematerialtransports fromhydrographic significantlysincepreindustrialtimes (e.g., since 1750), sectiondata,highdatadensities aregenerallyrequired.Water subtraction of theanthropogenic CO2transportfromthetotal chemistrydataare availablefor a maximumof ~36 discrete transportmeasuredin the early 1990s providesan estimate depthsper station(> 100 m resolution),comparedto the ~2 of the inorganiccarbontransportwhich occurredduring dbarresolutionof the conductivity-temperature-depth (CTD) preindustrialtimes. In a separatepaper,we will discuss the profiler data from which geostrophicwater transportsare implicationsof the anthropogenic CO2 transportfor the estimated.Althoughnutrients,oxygen,andtheCTD dataare storageof anthropogenic CO2 in theNorthAtlanticOcean. availablefor almostall of thesebottlesamples,TCO2 measurements are notavailablefor every stationandeverybottle depth because of analysistime limitations.To increasethe 2. Data and Data Interpolation TCO2 data densityprior to making the inorganiccarbon The mainhydrographic sections usedfor thecalculation of transportcalculations, someformof interpolation or mapping the meridionaltransportof inorganiccarbon in the South schemeis required. Recently, Goyet et al. [1995] demonAtlanticweretheWOCE sections A8 (11øS),A9 (- 19øS)and stratedthepotentialof quadraticspatialinterpolation of sparse A10 (~30øS). Thesesections wereoccupiedby theInstitutf•r TCO2 data. As they noted,however,consideration of nearprofilesof "mastervariables"suchas T, S, and Meereskunde,Kiel, Germany, on boardthe researchvessel continuous Meteor between 1991 and 1994 [Siedler and Zenk, 1992; 02 shouldallowsignificant improvement overspatialinterpoSiedler et al. , 1993; Zenk and Mi•ller, 1995; Siedler et al. , lationschemes.Previousstudieswhichusedmultiplelinear withsuchvariablesto interpolate TCO2 ontobottle 1996]. The totaldissolved inorganiccarbon(TCO2) values regression [Johnsonet al., 1995, 1998] were determinedusingsingle- data were performedon basinscales[Breweret al., 1995; ß 5øS 10øS 20øS 30øS 35øS 55øW50øW 40 øW 30øW 20øW 10øW OøE 10øE 20øE Figure 1. Map of the SouthAtlanticOceanshowing thegeographical positions of thegeostrophic velocities usedin the transportcalculations. Each star represents a locationmidwaybetweentwo adjacentstations.Pleasenote that there are overlapping sections at 11øSandtowardtheeastern endof 30øS(eastof 5øW).Themeridional sections areusedonlywiththe conservation constraints in the inversemodel(seeappendix,constraintB). HOLFORT ETAL.'SOUTH ATLANTIC CO2TRANSPORT 481 Table 1. SectionData Usedfor thisStudy Cruise WOCE Location (s) Date Section Bottles TCO2 Tripped Samples Institutions Meteor 28 A8 April1994 11øS 3840 1540 IfM / BNL Meteor 15 A9 February i99i i9øS 3290 690 hqM/ BNL Meteor 22 A10 January 1993 30øS 3420 1420 IfM / BNL Oceanus 133 - March1983 11øS, 23øS 3690 SAVE Leg3 - February 1988 25øS, 0øW 2560 0 1620 WHOI SIO/LDEO SAVE Leg4 December 1988 30øS 2290 1040 SIO/LDEO "Bottles Tripped" refers toallwater samples collected during thecruise, ofwhich only asubset were analyzed forTCO 2.Note that theOceanus sections werenotsampled forTCO2: values wereinterpolated using themethod described inthetextonthebasis of adjacent section data. These sections were used because thetransport estimates are highly contingent onthemass transport fields: hence weused allavailable sections. Abbreviations areasfollows: IfM, Institute furMeereskunde, Kiel;BNL,Brookhaven National Laboratory; WHOI, Woods Hole Oceanographic Institution; SIO,Scripps Institute ofOceanography; andLDEO, Lamont-Doherty EarthObservatory. Dashindicates non-WOCE cruises. Wallace,1995].Theresiduals forsuchlarge-scale regressions rationalefor the choiceof thesepredictorvariablesis given vary systematically with geographic position and depth by Breweret al. [1995] and Wallace[1995]. Figure2 presents a composite plotof theresiduals for theTCO2 data [Wallace, 1995]. for oneof thezonalsections We therefore developed a three-step procedure to mapthe derivedfrom theseregressions sparse TCO2 bottledataontotheCTD datausingmultiple usedin this study.The standarddeviationof the residuals linearregression. Ourapproach involved thedevelopment of fromthesefitsfor all TCO2 datais 3 •mol kg-1,witha for largerresiduals withintheupper200m (Figure localizedmultiparameter linearregressions from all the tendency 2) [cf. Breweret al., 1995].Thenow"complete" bottledata thenutrients domaincenteredaroundthe position(latitude,longitude, setfor TCO2 wasusedin step2 to interpolate For every pressure, ordepth) atwhich avalue wastobecalculated. A andTCO2 fieldsontothe CTD measurements. CTD depth (2 dbar resolution), a regression with independent horizontaldomainextentof 2øx 2ø waschosento providean adequate number of samples foreachdomain in mostcases. variablesAOU, $, !9, andao wasperformedusingthebottle availablebottle data which were locatedwithin a specified Thevertical extent oftheregression domains (2 x APreg ) was data that were located within the local domain (as defined variedwith pressure, starting at + 150 dbararoundthe above)centeredon theparticularCTD depth.Finally,the equations wereappliedto theCTD profile desired position close totheseasurface andincreasing with localregression depthaccordingto APreg = 150+ 0.1*P (1) whereP is the pressure.A variableverticalextentwas datato estimate a TCO2 concentration for eachdepthin the CTD record.Usingthisthree-step method,we attainedan effective TCO2 datadensity of -•50kmhorizontal resolution alongthe cruisetracksand2-10 dbarin the vertical(the relativelyslowresponse of theoxygensensor likelyreduces the effective vertical resolutionof the CTD-O 2 measure- required because of thestronger vertical gradients foundin ments).A similarapproachhasbeenrecentlyproposed by theupperocean andthelowerresolution ofthebottle dataat GoyetandDavis [ 1997];theprincipaldifferenceis ouruseof greater depths. In cases wheretherewereinsufficient bottle geographically localizedregressions throughout the water datatoconstrain a regression fit fora domain of thissize,the size of the domainwas doubled.This was necessaryfor the Oceanuscruises,on which no TCO2 measurements were column. madeandfor whichthe mappingontothe bottleandCTD datareliedon regression equations derivedfromdatacol- 3. lectedfromadjacent cruise tracks.A 5ø x 5ø areawasused The methodfor calculatingthe CO2 transportwassimilar to thatusedpreviouslyfor the calculation of heatandsalt transports andwasbasedon techniques introduced by Bryan [1962], Wunsch[1978],Roemmich [1980],etc.for theNorth Atlantic.Closelyrelatedcalculations usingSouthAtlanticdata have been performedby Fu [1981], MacDonald[1993], Holfort [1994], Saunders andKing [1995]andJ. Holfortet al. (The heattransportin the SouthAtlanticOcean,manuscriptin preparation, 1998;hereinafter referredto Holfortet for theMeteor15 (19øS)cruise for whichthedatadensity was relatively sparse. In step1, we first usedmultiplelinearregression to calculate thefew missing nutrientdatawithinthebottledata sets.We subsequently estimated TCO2 valuesfor all of the bottlesamples for whichtherewasnoTCO2 analysis. The latter was achievedusingregressions with the following independent parameters: apparent oxygen utilization (AOU), NO3,SiO4,salinity (S),andpotential temperature (O).A Method a!., manuscritxin preparation1998). In particular,Holfort 482 HOLFORT ET AL.' SOUTH ATLANTIC CO2TRANSPORT [1994] and Holfort et al. (manuscriptin preparation,1998) presentedcalculations of heat and freshwatertransports for the same sectionsdiscussed here, togetherwith a more transportacrossthe section.The compensating velocitywas defined as Af0 deraileddiscussionof the calculationprocedures.It is worth noting that the heat transportscalculatedfrom the sections discussed here agreevery well with thoseestimatedby other workersfor similarlatituderanges.Thesehavebeensummarized and placed in a global and historical context by - EK Am -H ¾'comp = AfO MacDonald and Wunsch [1996]. Am -H We usetheterminorganiccarbontransport(Tc) for thenet transportof TCO2acrossan oceansection,thatis, rc=f f v. Am -H rco2aaz (5) (2) The Tc was calculatedby integratingthe productof total Thedepth-independent or "barotropic" velocity•N• was specifiedon the basisof an assumed level of no motion (LNM). For the simplest LNM choice considered, the "baroclinic" case,thisbarotropiccomponent wasspecified to be zero for each stationpair (see (4)). However, we employed34 additionalapproaches to specifyinga level of no motion[Holfort,1994].Briefly,theseapproaches included(1) inorganic carbon (TCO2)withthein-situdensity (Ps, r,P)and the velocity (v) that is orthogonalto the section,from the American(Am) to the African (Af) continentover the entire water column(i.e., to the bottomdepth, - H). Becausewe haddiscretestationspacing,the integralwasreplacedwith a sum over station pairs. It was assumedthat the oceanis in geostrophic balanceexceptfor a directlywind-drivenEkman Residual (lamol kg-1) -20 -10 0 10 20 0 10 20 velocity.Two alternativeapproaches to definingthevelocity fieldwereused:a level-of-no-motion approach andaninverse model approach. 1000 - 3.1. Level-of-No-Motion Approach For computational reasons associatedwith different constraintsused for the calculationof heat tramport (which assumeszero net mass transport across a section) and salt/carbon/mass transport(in whicha nonzerosalttransport constraintis invoked),thevelocityfield wasbrokendowninto several components.This decompositionalso facilitated sensitivitystudiesthat examinedsourcesof uncertaintyin the inorganic carbon transport estimates.The velocity at any 2000II 3000 - location on a section was II v(x,z) =v/(x,z) +[v•r(x,z) - •r ] -¾'comp 4000 - 5000 - (3) - ¾'comp + ¾ where v'(x,z) is the "baroclinic"velocity which can be calculateddirectlyfrom thehydrographic dataandis defined so that for eachstationpair 6000 0 f -H az=o (4) andve•:is theEkmanlayervelocity for eachstation pair, whichwasassumed tobedistributed evenlythrough theupper 50 m of the water column and to be zero below that. It was -20 -10 Figure 2. The differencebetweentheTCO2concentration as measuredon water samplescollectedduringMeteor 15 (19 oS) and that predicted using 'the multiple linear regression approachdescribedin thetextfor step1 of thedatainterpolation procedure.For this cruise, missingnutrientdata were first interpolatedusinga multiple linear regressionusinga calculatedfrom HellermannandRosenstein's [1983] climatological averagewind stressdata. The integratedEkman transportoverthe entiresectionwasthenmasscompensated horizontal domainof 2 øx2 o. Therelativelysparse TCO2data by a spatiallyuniform,depth-independent velocityin order setfor thiscruiserequireda largerinterpolation domainof 5 o that this componentof the transportgave zero net mass x5 ø. HOLFORT ET AL.' SOUTHATLANTICCO2TRANSPORT useof a fixed horizontalpressurelevel for the entiresection (we usedlevels which were varied systematicallyevery 400 dbarbetweenthe surfaceand 5200 m); (2) selectionof certain key isopycnals andisotherms; and(3) useof a geographically variable level of no motion derived from a water mass analysis.Oneparticularspecification of thistypewasusedfor severalsubsequent calculations andis referredto as LNM-2. Detailsconcerningthis specificationare givenby Holfort et nfo 483 nfo -S Am -H Am -H Af O Am -H Af O al. (manuscript inpreparation, 1998).The ½•vM fieldderived from the level-of-no-motionchoicewassubsequently adjusted in orderthattherebe zero netmasstransportacrossthe entire sectionarising from this componentof the transport.This (7) + f f•œNM.S.@s,T,? dxdz "compensating" velocitywasdefinedas Am -H Af0 - LNM Af0 Am -H V'cømp = AfO (6) Am -H As with the Ekman transportmasscompensation, this adjustmentvelocity was distributeduniformly acrossthe entire section.Note that the barotropicmass transport compensation requiredto satisfythezero-net-mass transport constraintcould be quite large. However, when this was distributed uniformlyacrossanentiresection,thelargecrosssectionalarea involved(sectionN5000km wide by N5 km deep) implied that the corresponding barotropicvelocity adjustments were too smallto be directlymeasurable(of the where •s is the salttransportthroughthe BeringStrait. 3.2. Inverse Model Approach In order to reduce the uncertainties and avoid some arbitrariness associated with the choice of a level of no motion,we alsodevelopedan inversemodelusingadditional constraints to assistwith the specification of the barotropic velocityfield. Note thatwith thisapproach,we solvedfor the entirebarotropicvelocityfor eachstationpair, andtherewas orderof 0.001 m s-•). no artificialdivisionintothethreedistinctbarotropiccompoFinally, an extra, spatiallyuniform,barotropicvelocity nents used with the level-of-no-motionapproach.In the component, 3-s, wasspecified in orderto forcethenetsalt inverseanalysis,the constraintsare usedto set up a linear transport acrossthe sectionto matchthenettransportof salt equationsystemor "model"for the unknownvelocities,and whichenterstheAtlanticthroughtheBeringStrait.Evapora- the systemwas thensolvedusingsingularvaluedecomposition, precipitation,andrunoff,because theyinvolvefreshwa- tion. This methodwas usedby Wunsch[1978] in the North ter, have a negligible effect on the mass of salt contained Atlantic, and calculationsusing South Atlantic data were withintheocean(eventhoughsuchprocesses alterthesalinity performedby Fu [1981], MacDonald [1993], Holfort [1994] via dilutionand concentration). Therefore,if a steadystate and Holfort et al. (manuscriptin preparation,1998). We refer the readerto thesepublications for detaileddescriptions salt contentof the Atlantic Oceanis beingmaintained,the amount of salt which enters the Atlantic Ocean to the north of the inverse method; a listing of the constraintsthat we throughthe Bering Strait must be balancedby the same employedis given in the appendix. In performingthe inverseanalysis,solutionsbasedupon amountof saltleavingthe Atlanticto the south.Given a mass transport of 0.8 x 109kgs-i [Coachman andAagaard,1988] many differentmodelswere explored;thesedifferentmodels anda meansalinityof 32.5, thesalttransport throughBering included different formulationsfor the weighting of the Straitis approximately 26.7 x 106kg s-1. This mustbe additionalconstraintsand data as well as the three separate variantsof a phosphate transportconstraint(seeappendix). balancedby the meridionalsalt transportacrossour zonal sections in the SouthAtlantic.While therequiredcomponent The sensitivityof the solutionsto the initial specification of a of thebarotropictransport isvery importantfor theoverallnet level of no motion was also explored.The goal of suchan carbon transport across anoceansection(upto 1 x 109kgs-1, approachis to find a setof solutionsthatis consistent with the carrying upto2000kmols-i;seebelow),it does notnecessar- constraintsandthe datato within a specifiedstringency.An ily contributeto a carbon tramport divergencebetween exact solutionto the imposedconstraintsis generallynot sections. As a result,it hasoftennotbeenexplicitlyincluded desired because the constraints are themselves not an exact of the ocean'sbehaviorandhaveuncertainties, in prior calculations of the carbontransport.Someconfusion representation has arisen in the literature as a result of this distinction, as do the data. At high matrix rank, which corresponds to a however(see section7). This "extra"barotropictransport stringent specification of the constraints, sucherrorsand component,whichwas assumedto be distributeduniformly uncertaintiestendtb be magnified,resultingin tinreasonably acrossan entire section,resultedin a net masstransport high andvariablebarotropicvelocities.At low matrix rank, across eachof thesections. Theassociated velocity,3-s,was the solutionwill be out of the range of the uncertaintyof the calculated from data and/or constraints, and the solution therefore can be 484 HOLFORT ET AL.' SOUTH ATLANTIC CO2 TRANSPORT judgedincorrect.Our goalwasto f'mda setof solutions that sophisticated LNM choice(LNM-2) gavea somewhat larger satisfiedthe constraintswithin reasonableerror margins.It shouldthereforebe rememberedthat there is no uniqueor "correct" solution, but there is rather a range of different solutionsthat dependon the rank and the particularsof the model including its initial state and factors such as the weightingandscalingof thevariousequations andunknowns. valuefor Tc of-2345 kmols4 (Table3). A difference betweenthisvalueand Tc using the "baroclinic"choiceof LNM (Table3) of-442 kmols-1reflectsa smalladditional barotropic contribution to thenetcarbontransport (TcmU). Henceat 11øS,T•c andT••:arebothsignificant butcanbe viewedasroughlybalancingeachothersothatthe net carbon transportTc is carriedprimarily by the net masstransport 4. Results across thesection (TSc). and Uncertainties At 30øS, the Ekman masstransportis muchsmaller,and 4.1. Overview of the Inorganic Carbon Transport T••:isa negligible + 35kmols-• ofinorganic carbon (northward).At thislatitude, theTc swas(-0.53x 109kgs'• x 2189 In order to set the contextfor subsequent discussionof •mol kg4) or -1160 kmol s'• of inorganic carbon.The uncertaintiesin the carbon transport estimates,we first baroclinic component, T•c, wastherefore (-3758- 35 + summarizetheresultsfrom our analysisof thedata.Through- 1160)or -2633kmols4. Useof themoresophisticated LNM out we use the conventionthat positive numbersrefer to (LNM-2, Table3) reducedtheoverallTcvalueto -2720kmol northwardtransport.In the following,T refersto transport s4 reflecting an additional barotropic component (T• M)of estimates (with Tc referringto carbon,and T• referringto + 1038kmols-•. Henceatthislatitude, themass-compensated mass). C denotesthe concentrationof total dissolvedinorEkman(T•) andthe combined barotropic (r• M + Tc s) ganiccarbon.Superscripts areusedto distinguish thedifferent contributionsto the net carbontransportwere both small so components of the transport(seesection3)' EK refersto the thatTc wasdominated by thebarocliniccirculation compomass-compensated Ekman component,BC the baroclinic nent(T•C). component,S refers to the net mass transportcomponent With this approximatebreakdownin mind, we now required to satisfy the salt transportconstraint,and LNM examinethe uncertaintyin the carbontransportwhicharises refers to the mass-compensated barotropic component from uncertaintiesin the specification of the differentformal associatedwith the choiceof a level of no motion. Angled components of the transport(mass-compensated Ekman, brackets(() denotean averagedquantity.Carbontransports baroclinic,netmasstransportaswell asthederailedspecificaarereported in kmols'• (1 Gt C yr4 = 1 Pg C yr4 = 2642 tion of the barotropicvelocityfield). kmols'•). The overall best estimatefor the meridionalinorganic carbontransport(Tc) in the SouthAtlanticOcean,between 11øSand 30øS,wasof the orderof-2150 lanol s4 (i.e., southward). This estimate, which was derived from the inversemodelingapproach,appliesto Tc at the time of the measurements: theearly-to-middle1990s.Usingthelevel-ofno-motionapproachdescribedin section3, we canview the Tc acrosseachsectionasbeingcarriedby thefourmainmass transport components listedabove:/•c, /ea:,/•u, andif. Usingthe resultsof the analysespresentedin Tables2, 3, and 4, it is possibleto distinguishtheir relative contributions to rc. Table 2 indicatesthatat 11øS,therewasa largesouthward Ekmanmasstransport (T•e•:)persewithintheupper50 m of 4.2. Ekman Transport For the purposeof illustratingpotentialerrorsassociated withspecification of theEkmantransport,weusethelevel-ofno-motionapproachwhich views the mass-compensating transportasa zonallyuniformbarotropicmasstransport(see section3). We note below that this approachmay give a worst-caseview of the transportuncertaintiescomparedto, for example,theinversemodelapproach.The mass-compensatedEkmancomponent of the totalcarbontransportacross thesection (T••:) canbeapproximated by .(<C•/•>-<C>) W = r;/EK (8) the water column, carryingthe TCO2 concentration of this upperlayer(Ce•:).The(southward) carbon transport in this layerwastherefore verylarge(-12x 109kgs4 x 2055•mol kg4) or -24,660lanols'•. Howeverthisis offsetby the thatin ourfullcalculations, weresolved Ce•:andT•e•:foreach northward mass-compensating barotropic transportwhich carries the higher concentrationcharacteristicof the water columnaverage( < C > ). The compensating transportcarded stationpair.) Table2 presents valuesfor the variablesin (8) for the six sectionsusedin this study.As notedearlier, the net Ekman (+ 12 x 109 kg S'1X 2196•mol kg-•) or +26,352kmols4. Hencethe net carbontransportattributableto thiscomponent (Tb) wasdirected northward at + 1700kmols-•. Table 3 presentsresultscalculatedusingthe "baroclinic" ß_• - 1 •UD where< C> and< Ce•:>areaverage TCO2concentrations of theentiresectionandtheEkmanlayer, respectively.(Note contribution is approximately + 1700kmols-1at 11øSand close to zero at 30øS. The concentrationdifferencein (8) remains atabout140ttmolkg4 forall of thesections, andthe variation in Tc ex fromonesection to another is therefore almostexclusivelya timarianof t2neE'._•2v. mnmasstramport s4 at 11øSincludes thecontribution fromTc exdiscussed inthe (T•eg, seeTable2). Thereisconsiderable uncertaintyconcerning themagnitude previous paragraph, aswellasfromTc s. Thelattercarries (-0.92 x 109kg s-• x 2196 •mol kg-•) or -2020kmol s4 of T•eg arisingfromuncertainties in windstress data.For (Tables 2 and 4). Hence the transportof carbonassociated comparison,thelargestdifferencesof meanzonalwind stress withthebaroclinic structure ofthevelocity field(7•c)at11øS in the North Atlantic between the data sets of Hellermann and is (-1903- 1700+2020) or -1583kmols-•. Useof a more Rosenstein[1983] and Isemer and Hasse [1985] are of the HOLFORTET AL.' SOUTHATLANTICCO:TRANSPORT 485 Table2. Values oftheMass-Compensated Ekman MassTransport (T•K), theMeanTCO2Concentration fortheEkman Layer( < C> eK)andtheEntireSection ( < C> ), andtheNetCarbon Transport Associated withtheMass-Compensated Ekman Transport (Tong) forSections atVarious Latitudes. Cruise Latitude, T •s: deg S 109•g's 'l < C> •s:, < C> , /zmol kg 'l Tc•c /zmol kg 'l kmol •-I Meteor 28 II -11.8 2057 2196 + 16,/.. Oceanus 133 11 -12.0 2053 2196 4-1744. Meteor 15 19 -6.3 2071 2193 +785. Oceanus 133 23 -3.8 2057 2188 + 503. SAVE Leg 3/4 25 -1.5 2047 2190 4-231. 30 -0.4 2046 2189 +35. Meteor 22 Transports are positivenorthward. Table 3. Sensitivity of CarbonTransportEstimates to Uncertainties in EkmanTransportandto the SpatialResolution of the Section Data. ModelParameters Choice ofLevelof NoMotion Transports (Tc), kmols'l Resolution HEK m zadjust Meteor 11øS28 Vertical Meteor 19øS 15 Meteor 30øS 22 Horizontal Uncertaintiesin EkrnanTransport Baroclinic 50 - 1 1 -1903 -2741 -3758 Baroclinlc 20 - 1 1 - 1903 -2734 -3761 Baroclinic 100 - 1 1 -2045 -2793 -3762 Baroclinic 50 -0.2 1 1 -989 -2223 -3417 Barochnlc 50 +0.2 1 1 -2816 -3259 -4100 Barocllmc 50 SOC 1 1 -2594 -3026 -3793 SpatialResolutionof the SectionData Bar oclinic 50 - 1 1 - 1903 -2741 -3758 Baroclinic 50 - 1 0.5 -2034 -2862 -3805 LNM-2 50 - 1 1 -2345 -2317 -2720 LNM-2 50 - 1 0.5 -2423 -2562 -3162 LNM-2 50 - NODC 1 -2186 -2932 -2884 LNM-2 50 - NODC 0.5 -2151 -2956 -3620 LNM-2 50 - Met25 1 -2545 -2865 -2838 LNM-2 50 - Met25 0.5 -2574 -3110 -3302 LNM-2 50 - Met25 0.33 -3030 -3642 3545 HEI•istheassumed depth oftheEkman layer;"zadjust" refers toadjustments (dynes cm':)totheHellermann andRosenstein [1983] wind stressdata; "SOC" refersto the altenativewind stressclimatologyfrom the Southampton Oceanography Center (seetext); horizontalresolutions of 0.5 and0.33 referto subsampling of sections usingeverysecond andeverythirdstation,respectively. Results for three differentverticalresolutions are presented:1 (full 2 m resolution),NODC (standardbottle depthsof the National Oceanographic DataCenterasof 1988),andMet25(standard bottledepths usedduringthehistoric Meteorexpeditions of 1925and 1927). 486 HOLFORT ETAL.' SOUTHATLANTICCO2TRANSPORT orderof 30% (i.e., 0.3 dyncm'2at 15øN).Onthebasisof the NorthAtlanticcomparison, we estimated theuncertainty in T• using anassumed errorinwindstress of + 0.2dyncm'2. Examplesof the sensitivityof the totalcarbontransportto this uncertaintyfor three of the sectionsare presentedin Table3. T•e•: is a fimcfion of thewindstress 0:) andthe Coriolisparameter: months, andweusedclimatological average windstress data to estimatethe Ekmantransport.Our calculations therefore implicitlyignore(or average)seasonal factorsthat might affect the carbon transport.Using data collectedat the BermudaAtlanticTime SeriesSiteas an example,seasonal changes of TCO2concentration of theorderof 40 •tmolkg'l canoccurin theoligotrophic oceanwiththisvariabilitymost pronouncedin the upper 50 m - 75 m of the water column S (9) [Bateset al., 1996]. At 11øS, wherethe surfacelayer mass transport is strongest (about10x 109kg s'l) dueto strong Ekmantransport,variabilityof thismagnitude couldtranslate Hence for a given uncertaintyin wind stress(r), the uncertaintyin Tc is muchgreaterat 11øS (+50% or 900 into seasonal Tc variabilityof the orderof 400 kmol s'l. Variability at the higher-latitudesectionswould be smaller than this because of the smaller Ekman contribution to the kmols'i, Table3) thanat30øS(+ 10%or400kmols'l, Table transport. However,in thissimplecalculation weareignoring 3) asa resultof thevariationof theCoriolisparameterfwith possible seasonal covariance of upperlayerTCO2concentralatitude.Useof an alternative windstressclimatology from tionswith theEkmantransportaswell asanytimedepend- the Southampton Oceanography Center[Joseyet al., 1996; enceof thenon-Ekmantransport.Obviously,we havea lack (seealsowww.soc.soton.ac.uk/JRD/MET/fluxclimatology. of data with which to reliably assesssuchseasonaleffects, html) S. A. Joseyet al., New insightsinto the oceanheat andthiswill be an importantissuefor furtherdatacollection budgetclosureproblemfromanalysisof theSOCair-seaflux andmodelingstudies. climatology, submitted to Journalof Climate,1998]givesa southward inorganiccarbontransportthatis 691 kmol s-1 greaterat 11øS,withthedifference decreasing to 35kmols'l at 30øS (Table 3). 4.4. Data Resolution It is worthnotingthatwhereascarbontransports basedon thelevel-of-no-motion approach appearverysemifiveto wind stressdatauncertainty, thismaybe anartifactof thesimple approach usedtospecifythecompensating barotropic velocity field. The sensitivityto variationsof wind stressof the full inverse modelresults wasconsiderably smaller.Forexample, at 11øS,theinversemodelcarbontransport variedby only about200-300 kmol s-1for the variouswind stressscenarios given in Table 3. The reasonfor the reducedsensitivity presumablylies in the details of how the inversemodel distributesthe transportrequiredto masscompensate the Elanantransport. Anotherpossiblesourceof uncertaintyis the choiceof secondand every third station(resolutionfactorsof 0.5 and 0.33, respectively).We alsoexaminedthe effectsof vertical Elananlayerdepth(/_/e•:), whichinfluences thevalue of data resolution and combined vertical and horizontal resolu- Accurate calculations of Tc fromzonalsections require veryclosestation spacing, because sums overstation pairsare substituted fortheintegralin (2). We cannot directlyestimate the error associated with thissubstitution. However,with a mean stationspacingof -•50 km, generallycloserat the boundariesandwider in the oceaninterior, themain scalesof motionsshouldbe resolvedby our data(thiswasnot neces- sarilythe casewith someearlierstudies;seesection7). In order to providean indicationof the possibleerror, we "artificially"'subsampled threeof thesections usingevery (C e•:) in (3). Normally, we usethemeanTCO2of the tion by subsamplingthe full sectiondata at standardor upper50 dbarof thewatercolumnastheTCO2concentration traditionalbottlesamplingdepths.Two variantsof standard for the Elanan flow; however, we also made calculations depths wereused:thestandard bottledepths usedby theU.S. National OceanographicData Center as of 1988 and the using meansof the upper 20 and 100 dbar to assessthe sensitivityof the carbontransportto thischoice.The results standarddepthsemployedon the historicMeteor cruisesof presented in Table 3 suggestthat the overalluncertainty 1925 and 1927. In all thesereducedresolutioncases, the associated (vithchoosing different Elanandepths is small entirevelocityfield wasrecalculatedbasedon the "artificial" ( • 100kmols'l) because thechange of ( Ce•:) withvarying dataset, andthe propertytransportswere recalculated. Ekmandepthis small.Again,theeffectis negligible at 30øS dueto the low Elananmasstransportat thatlatitude.Similarly,theuncertainties associated withanuncertainty of about The resultsare also presentedin Table 3. The effect of halvedhorizontalresolutionappearsto be relativelyminor layer (seeFigure 2) are small. decreasedvertical resolutionalso varied betweensections, (change in Tcof -•160kmols'l),although a largerdifference 5 t•molkg'l inthemapping oftheTCO2values inthesurface at 30øSwasobservedusingtheLNM-2 model.The effectof largeeffect(-•600kmols-1)being observed Thepotentially largeuncertainties associated withspecify- withaparticularly at 19øSbut with effectsat theothersections beingonly 100ing the Elanantransportimply that an importanttaskfor it canbeseen that,depending onthe improvingtransportcalculations, not only for innrg•nic 200kmols'l. In general, section,bothhorizontalandverticaldataresolution changes carbonbut also for heat, nutrients,etc., is to obtainbetter can affect the inorganiccarbontransportestimates.Table 3 estimates of thewindstress,particularlyat low latitudes. shows that the combined effect of decreased vertical and 4.3. Seasonality horizontalresolutions canbe quitesignificant, withchanges of 700-1300lanols-1 (againdependent on theparticular The sectionswere collectedprimarilyduringsummer sectionexamined). 487 HOLFORT ET AL.' SOUTH ATLANTIC CO2TRANSPORT AtlanticDeep Water (02 > 37.07 and 04 < 45.92) flowed southward; (3) theflow of upperNorthAtlanticDeepWater 4.5. Barotropic Component A measureof theuncemintyin Tc thatcanariseasa result of uncert•ntyin thespecification of thebarotropiccomponent of thevelocitycanbe obtainedfrom exploringthe sensitivity of Tc estimates to differentchoicesof a level of no motion (LNM). As describedin section3, we employed35 different LNM choices. Some of the choices resulted in meridional overturning circulationsthat are incompatiblewith our knowledgeof theoceancirculation.We thereforedefinedthe circulationassociatedwith a particularLNM choiceto be "valid"if the followingcriteriaconcerningthe circulationof major water masseswere met: (1) the flow of Antarctic BottomWater (o4 > 45.92) wasnorthward;(2) lowerNorth waslessthan26 x 109kgs-l;and(5) theAntarctic IntermediateWater (26.8 < o0 < 27.4) flowednorthward. Thefrequencydistributions of Tc estimates resultingfrom theLNM choices thatgave"valid"circulations arepresented for thedifferentsections as solidbarsin Figure3; theopen bars representtransport estimatesassociatedwith LNM choicesthat gave "invalid"circulations.The meanvaluesof Tc for the "valid"choices(Table4) lie between~2000 kmol s-1and~3500kmols'1to thesouth,andthemedianvaluesdo not differ very much from thesemean values.The standard Meteor 22; 30øS Meteor 15; 19øS Meteor 28; 11øS I 16 14 o (02> 36.9and02< 37.07)wasatleast5 x 109kgs'1tothe south;(4) the totaltransportof North AtlanticDeepWater I Valid [•1 Invalid I I I I I I I I 16 C A 12 12 lO 10 8 ß- 6 6 4 4 I ...... 0 0 0 I ...... 0 0 0 0 I 0 0 0 0 0 0 .... I 0 I 0 0 0 FI o 0 0 0 TCO2Transport (kmols'•)Southward Oceanus 11øS SAVE 25øS Oceanus 23øS ' 16 ' I I i i I • 16 Valid D 14 14 Invalid 12 12 o >' lO 10 8 ß • 6 6 4 4 2 0 0 n ....... o,, 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 ! I-I 0 , o o o ,•. TCO2Transport (kmols'•)Southward Figure3. Frequency distributions oftheTc (inkmols'• onthex axis)forthesixdifferent zonalsections based onthelevel-ofno-motionapproach.Solidbarsrepresent"valid"choicesof a levelof no motion;openbarsrepresentlevelsof no motionthat resultedin an "invalid"circulation(seetext for details).For thesesolutions,anydifferencebetweenthe salttransportacross the sectionrequiredto balancethe salttransportthroughBeringStraitandthecalculated salttransportbasedupona choiceof a levelof no motionhasbeencompensated for by an additional,horizontallyuniform,barotropicvelocity. 488 HOLFORT ET AL.' SOUTH ATLANTIC CO2TRANSPORT deviations rangebetween200 and650 kmols'• withthetotal range(maximum-minimum) for an individualsectionbeing ~1600kmols-•. Hencespecification of a levelof nomotion can be a very significantsourceof uncertaintyfor the Tc estimates. It wasin orderto reducethisuncertainty thatwe developed the inversemodel analysisdescribedin section3. Solutions baseduponmanydifferentinversemodelswereexplored,and the sensitivityof the solutions to the initial specification of a level of no motion was also explored.A fairly consistent pictureemerged:At low matrixrank,thesolutions werestill influencedsignificantlyby the particularchoiceof initial LNM; however, at ranks of > 80 this influencelargely disappeared.At ranks > 200, the exact natureof the constraintsbecameimportant,andthe effectsof modelanddata un.certaintieswere amplified so that solutionsfor different modelsdivergedsignificantly.There was a relativelystable regionof carbontransportversusrank in the range 135 < rank < 165. We chosethisrangeof solutionsasthebasisfor the transportestimatespresentedin Table 4. This table presentsmeantransportsfrom the inverseanalysistogether withestimates of thestandard deviationacrossmultiplemodel formulationsfor thisrangeof ranks. In general, the transportsacross the various sections derivedfrom the inversemodelapproachare lower thanthe meanvaluesderivedfrom the level-of-no-motion approach, particularlytowardthe south. Examinationof the solutions suggestedthat for the southernmost sections,it was the phosphate transportconstraint thatwasa principalreasonfor the reducedtransportestimaterelativeto thosederivedusing the level-of-no-motion approach. The inorganiccarbon transport wasquasi-linearly relatedto thephosphate tramport with a slope(/iTc //iTj•o4) of 110-170,whereTj•o4is the phosphatetramport. Overall, however, the uncertainty associatedwith the barotropicvelocity specificationwas reduced through useoftheinverse modelto ~200kmols-•for an individualsection,andtherewasalsoreducedvariability between the means for the individual sections. 4.6. Net Mass and Salt Transport In order to satisfythe constraintof a total southwardsalt transport of 26.7 x 106 kg s'• through the sections, our calculationsrequirea net southwardmeridionalmasstram- portof-0.96x 109kg s-1at 11øS,decreasing to -0.53x 109 kgs'• at30øS(seeTable4). Thefreshwater balance (evaporation,precipitation,andrunoff) inferredfrom thisestimate of the convergenceof the masstransportagreesvery well with the completelyindependent estimateof net freshwater fluxesfor this latituderange(0.42 x 109kg s'l) by BaumgartnerandReichel[1975]. As notedearlier,assigning thisnet masstransportto a zonallyuniformbarotropicflow givesan associated inorganiccarbontransportof about-2000 kmol s'• at 11øS and -1160 kmol s4 at 30øS due to this componentof the transportalone. The uncertaintyof the Tc due to mostof the sourcesof uncertaintyin thisnet masstransportis alreadyincorporated in therangesof solutions presented in Table4. Thesedidnot, however,considerthe uncertaintyin the specification of the salttransport throughBeringStrait(andthereforethroughthe SouthAtlanticsections). Anuncertainty of 0.2 x 109kgs-1or HOLFORT ET AL.' SOUTHATLANTICCO2TRANSPORT 25 % in the masstransportthroughBeringStraitresultsin an uncertainty of 6.5 x 106 kg s'l in thesalttransport which translatesinto an associatedadditionaluncertaintyof ~400 kmols4 forTcintheSouth Atlantic.(Notethatsuchanerror would be systematicacrossall SouthAtlantic sectionsand therefore would not affect carbon transport divergence estimates;seesection7.) 489 as a functionof the 19 separatedensityintervalsusedin the inversemodel. The corresponding densityintervalsare listed in the Figure 4 caption, and their depth distributionsare shownin sectionplots (Figure 5). A three-layersystemis evidentin themasstransportdistribution,withpredominantly northwardtransportin thelowerdensitylayers(ol < 32.00) which include the Antarctic Intermediate Water, southward transportin the middensitylayers(o• > 32.00 and o3 < 4.7. Structure of the Transport 41.53) which includethe North Atlantic Deep Water, and The quasi-vertical structureof the section-wide transports of mass,silicate,andTCO2 for theMeteor22 cruisealong 30øSare presentedfor onetypicalrealizationof the inverse analysisin Figures4a - 4c). Note that at this latitude,the Ekman transportis small and is includedin the uppermost densitylayer. In theseplots,the totalnet transport. is plotted northward transport in thehighest-density layers(o3 > 41.53) which include the lower Circumpolar Deep Water and Antarctic The carbontransport(Figure4c) appearsqualitatively very similarindicatingthatthe overalldetailsandmagnitudeof the inorganiccarbontransportare controlledto first order by Mass Transport (109 kgs'1) -6 -4 -2 i I 0 2 4 I I Bottom Water. Silicate Transport (kmol s'1) 6-200 -100 0 i .:,:.:.:.:.:.:.:.:.:,:.:.:.:.:.• '..3 ':3 .',.•1 :.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.3 ':;:;:;:;:;3 :,:.:.:.:,:,:,:.:,:.:.:,:.:.:,1 I;:;:;:;:;:;: lO i 300 400 i 0 I.:-:.:.:.:.:,:.:,:.:,:.:,:.:.: F:.:.:.:.:.:.:,:,:.:.:.:, lO ... :::::::::::::::::::::::::::::::::: 1:;:;:;:;5:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;: (1) i:;:;:;:;:;:;:;:;:;:;:;:;:;:;1;:;:;:, •;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:; i:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:,:.:.:.:.: t.:.:.:.:.:,:.: ß!: 200 I:;:;:;:;:;:;:;:, F:,:.:,:.:.:.:.:.:.:.:.:.:.:, •.:.:.:-:.:.:.:.:,:.:.:t:.:•:,:•: ß 100 i ;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;:;5:;:;:;:;:;:;:;$ i.:.:.:,:.:.:, 15 - 15 :.:.:.:.:.:4 o '!: o ;:;.;:;.;:;:;:;1;:;.;.;.;:;..;:;:;:• B I 20 :.:.:.:.:,:.:.:.:.:.:•.:.:.:.:.:4 I I I 5:;:;:;:;:;:;:;:;:;:;! .:.:.:.:.:.:,:.:,:.:.:.:,:.:.:.:.:,:,:,:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:. I I I ! 20 0 0 ß lO lO (• ß!: 15 15 '!: o o 20 20 -8000 -4000 0 4000 8000 InorganicCarbonTransport (kmol s'1) -20 0 20 40 60 80 100 Anthro. CO2 Transport (krnol s-1) Figure4. Verticalstructure of thetransport of (a)mass,(b)silicate, (c)TCO2,and(d)anthropogenic CO2at30øS(Meteor22) as derivedfrom a typicalinversemodelsolution.Northwardtransportis positive.The verticalaxis givesthe densitylayer numberfrom theinversemodel.The locationsof thelayersare shownin sectionplotsin Figure5 andweredefinedasfollows: 1, o0 < 26.60; 2, 26.60 < o0 < 26.80; 3, 26.80 < o0 < 27.00;4, 27.00 < o0 < 27.20; 5, 27.20 < o0 ando• < 32.00; 6, 32.00 < o• < 32.16;7, 32.16 < o• ando2 < 36.82;8, 36.82 < o2 < 36.92; 9, 36.92 < o2 < 36.97;10, 36.97 < o2 < 37.00; 11, 37.00 < o2 < 37.02; 12, 37.02 < o2 < 37.04; 13, 37.04 < o2 ando3 < 41.50; 14, 41.50 < o3 < 41.53; 15, 41.53 < o3 ando4 < 45.93; 16, 45.93 < o4 < 45.96; 17, 45.96 < o4 < 46.00; 18, 46.00 < o4 < 46.02; and19, o4 > 46.02 490 HOLFORTET AL.' SOUTHATLANTIC CO2TRANSPORT __ ooo •--- 32.00 • 1000:,/•;.v • 2.'7. U"-• ooo I ' 56.9'/• 2000 3000 ooo 4OOO 000 5000 000 .•.50 45.93 000 6OOO -30 -20 -10 0 -30 10 -2_0 -10 0 10 longitude longitude ½ Ob ....... , ......... ' ......... , ......... "&6:8"-.-:-• 1000 2000 3000 4OOO 5000 6000 -30 -20 -10 0 10 -40 longitude -30 -20 -10 0 10 longitude Figure5. Distribution of the19layersusedfor theinverse modelanalysis in fourof thesections usedfor thisstudy.(a) Oceanus 133at 11øS;(b)Meteor28 (WOCEA8) at 11øS;(c)Meteor15(WOCEA9) at 19øS;(d)Meteor22 (WOCEA10) .at30øS.In allcases, thelayernumber wasassigned sequentially fromsurface todeepaccording tothedensity criteria listed •n the captionto Figure 4. variability in the mass transport and that concentration variabilityplaysa smallerrole. Conversely,the influenceof verticalconcentration variabilityis very evidentin the caseof the silicatetransport(Figure4b), wherethe effectof silicate depletionin low-densitylayersandveryhighsilicateconcentrationsin the highest-density layer canbe seento affectthe overallpatternof the transportstrongly. transportshoweda strong,northwardflowing subsurface currentwhich corresponds well with the combinedacoustic dopplercurrentprofiler / CTD observations of $chottet al. [1995]. The solution also revealed a southward flow of AntarcticIntermediate Wateralongthewesternboundaryat 30øS,asobservedfrommooringsandfloatstudies[Boebelet al. , 1997]. As with the vertical structure,the horizontal structureof the inorganiccarbontransportis dependenton the mass transportdistribution.The horizontalmasstransportderived from the inversemodelanalysisis a complexsubjectwell beyondthe scopeof this paper:it is coveredin detailby Holfort[1994]andHolfortet al. (manuscript in preparation, 1998).However,theinferredcirculationpatternwasfoundto be broadlysimilar to that of Reid [1989], althoughsome differenceswere alsoevident.At the westernboundary,the southwardflowing Brasil Current at 30øS (the southern boundaryof the inversemodeldomain)was prescribedto 5. AnthropogenicCO2 and the Preindustrial Transport Theestimates of Tcpresented aboveareappropriate forthe periodof time duringwhichour TCO2 measurements were made, the early 1990s. However the concentrationof dis- solvedinorganic carbonin theoceanis increasing withtime asa resultof theuptakeof anthropogenic CO2 derivedfrom land use changesand fossilfuel burning.The near-surface concentration of TCO2atthislatituderangehasincreased by havea transport of 10 x 109kg s-• (based ondirectcurrent ~55/•molkg-• since preindustrial timesin response to the measurements). At 11øS (our northernboundary),the mass increase in atmospheric (andnear-surface ocean) pCO2from HOLFORTET AL.' SOUTHATLANTICCO2TRANSPORT ~275-280 •atm in 1700 [Neffel et al., 1985] to ~355-358 •atm for the time of the WOCE surveys[Conwayet al., 1994]. It is worth consideringwhatthe oceanictransportof inorganiccarbon within our study region would have been duringpreindustrial times.In doingthis,we implicitlyassume that oceancirculationpatternsand rateshave not changed significantlysince1750. (We notethatobservations of longterm changesof water propertiesat variouslocationsin the oceanassociated with changingatmosphericforcing [e.g., Dickson et al., 1997] may prove difficult to reconcilewith this assumption.) In orderto estimateTc for theperiodprior to theintroductionof anthropogenic CO2 (i.e., prior to about1750), we first estimated the anthropogenic CO2 contentof our TCO2 data. We used a variant of calculationprocedures,basedupon preformedCO2, whichwereintroduced byBrewer[1978]and 491 early1990s.Thetransport of anthropogenic CO2to thenorth arisesbecauseit is theupperoceanwaterswithinthislatitude rangethatcontainmostof theanthropogenic CO2 , andthese are moving predominantlynorthward(Figure 5d). Conversely,the southward returnflow of "older"deeperwater contains a muchlowerconcentration of anthropogenic CO2. This supplyof anthropogenic CO2 to the northernand tropicalAtlantic Ocean which is carried by the ocean's meridional overturning circulation isquantitatively significant relativeto theuptakeof anthropogenic CO2acrosstheair-sea interfacewithinthe NorthAtlantic.For example,Sarmiento et al. [1995]reporteda model-derived estimateof theuptake of anthropogenic CO2 dueto air-seagasexchangewithinthe northernAtlantic (18øS - 78øN) of-1000 kmol s-•. Our estimateof the transportof this propertynorthwardacross 20øSwithinthe oceanapproaches 50% of thisvalue.We will Chen and Millero [1979] and included the contribution of discussthe importanceof this meridionaltransportfor the nitrate remineralizationto alkalinitychanges[e.g., Brewer overallbudgetof anthropogenic CO2withintheNorthAtlantic and Goldman,1976]. It is beyondthe scopeof thispaperto elsewhere. discussthecalculationprocedurein detail;however,it shared Subtraction of theanthropogenic carbontransport fromTc severalsimilaritieswith an approachemployingtransient allowsestimationof the inorganiccarbontransportduring tracer data developedby Gruber et al. [1996]. The estima- preindustrial times(Table4). The transportacross10øSwas tionof theanthropogenic CO2concentration doesnotcontrib- about-2900kmol s-• (i.e., to the south),whichcan be ute greatlyto the overalluncertaintyof eitherthe anthropo- comparedwith the estimatedpreindustrialtramportinto the genicor preindustrialcarbontransportsrelativeto uncertain- Atlantic via the Bering Strait of about 1630 kmol s-• tiesassociated with thecirculationandmasstransports. [Lundbergand Haugan, 1996]. There thereforeappearsto The averagedepthprofileof anthropogenic CO2for each havebeena preindustrial (natural)divergence of theinorganic of the sections is presentedin Figure6. Very noticeablein all of theprofilesis a relativelyhighanduniformanthropogenic Anthropogenic CO2 (lamol kg-1) CO2 concentrationin the Antarctic IntermediateWater betweendepthsof 500 m and1000m. Overall,theanthropo0 10 20 30 40 50 60 I I I I genicCO2 concentrations in the water columnare highest towardthe south,closerto the origin of the AntarcticIntermediateWater[cf. Brewer,1978].Our analysisgivesa small averagebackground levelof anthropogenic CO2of 3-5 •mol kg4 throughout the watercolumnin eachsection.This lOOO nonzero background is consistentwith the detection of significantlevels of the anthropogenic halocarbonCC14 throughoutthe water column in the Brazil Basin and even in portions of the deep Angola Basin [Wallace et al., 1994; Roether and Putzka, 1996]. Carbon tetrachloride has been presentin the environmentin significantquantifiessince 1920-1930at whichtime the anthropogenic CO2 contentof thecoldsurfaceseawaterwhichventilates thedeepoceanwas 2OOO 3OOO already12-14•mol kg-•. Theanthropogenic CO2,having beenintroducedsinceabout1750,mayhavepenetrated farther into the deepoceanthanCC14andconsiderably fartherthan the CFCs and bomb tracers such as 3H and •4C. Hence the findingof CC14throughout mostof thelow-to middle-latitude SouthAtlanticbasinssuggests thatno portionof the Atlantic Oceancan safelybe assumedto be "free"of anthropogenic CO2 contamination. The anthropogenic CO2 distributions fromthe sectiondata were combinedwith the masstransportcalculations in order tocalculate thetransport of anthropogenic CO2. Theresulting transportof anthropogenic CO2wasconsistently northward, with a valuewhichrangedbetween400 and700 kmol s-• at 4OOO , 5000 11os ---=-- 19os --•-- 23os -•--30os .....* .... 11os (M28) ---',-'- 25os (SAVE) 6000 I I 11øSandwas300 kmols-• at 30øS. The anthropogenicFigure 6. Averagedepthprofileof anthropogenic CO2 for carbontransportspresentedin Table 4 imply thatthe south- thesections usedduringthisstudy.Eachprofilerepresents the wardinorganic carbon transport Tcwas~430kmols-•greater averageprofile for an entiresection. (Note thatmostindividacross20øSduringpreindustrial timesthanit wasduringthe ual profileshad bottomdepthsshallowerthan6000 m.) 492 HOLFORT ET AL.- SOUTH ATLANTIC CO2 TRANSPORT Bering 90 90N 10S 30S Bering Eq õ60 Strait Strait Antarctica 10s 30s 90N Bering 60S 1280' / 650 Eq • 60s Antarctica o 90N 10S 30S o o • • • 6OS g Unitsare kmol s4 (1 GT(C)yr4 = 2642kmols-') Figure7. Schematic summary of theregionalAtlanticOcean inorganic carbonbudget(10øSto 30øS)basedon thewater columntransport andstorage estimates. (a) thecontemporary budget,(b) thebudgetfor anthropogenic CO2 , and(c) the preindustrial budget.Thetransport andstorage unitsarekmol s4 . The number+240 in Figures7a and7b refersto the storageterm. In Figure7c, the budgetis extendedto the NorthAtlantic,andtheasterisked numberat thetopboundary refers to the combined effect of air-to-sea and rivefine inorganiccarbonfluxes. o o o • o carbontransportbetweenBeringStraitand10øSof the order of 1300kmols-1,whichwaslikelybalanced bynetair-to-sea transferof CO2 and riverineadditionsof inorganiccarbon within the North Atlantic and Arctic Oceans(seeFigure 7c). The potentialrole of organiccarbontransports mustalsobe considered in assessing thislarge-scaledivergence[Sarmiento et al., 1995]. 6. RegionalCarbon Budget(10oS - 30 oS) The carbontransportversuslatitudedataof Table4 canbe usedto examinethe regionalinorganiccarbonbudget.Figure 7 summarizes the followingdiscussion schematically for the contemporary, 1990sTCO2budget(Figure7a), theanthropo- tt• tt• 493 HOLFORT ET AL.' SOUTH ATLANTIC CO2TRANSPORT genic CO2 budget (Figure 7b),andthepreindustrial TCO2 section, weestimated theMtiDforanthropogenic CO2tobe 770 m for the region between10øS and 30øS. The uncerA regression of Tc versuslatitude(Figure8a)indicates that taintyof thisestimatemay be of the orderof 20% [cf. Gruber the contemporary(i.e., 1990s)inorganiccarbontransport et al., 1996]. Gammon et al. [1982] have shown that the tracerwith a seadecreased by 16.7 (_+19) kmols'• per degreetowardthe verticalprofile andMPD of a conservative surfaceconcentration thatis increasingexponentiallywith esouth,implyinga very slightconvergence of 330 (+380) kmol s'• between10øSand30øS(uncertainties are standard foldingtime of t yearsreachesa "transientsteadystate"after CO2, errors).Any contemporary convergence of inorganiccarbon (3 x t) years [Gammonet al., 1982]. Anthropogenic which has been increasingexponentiallysinceat least the canbebalancedby a sea-to-airflux of CO2,netproduction of organiccarbon,net dissolution of solidcarbonates, andnet mid-1800swith t = ~30 years, shouldby now have achieved storageof anthropogenic CO2. (The regionalriver discharge transientsteadystate. Hence the storageor the inventory asgivenbyPerryet al. [1996]is < 107kgs-• whichwould change of anthropogenicCO2 during the 1990s can be carrya negligibleinorganiccarbontransportof < 10 kmol estimateddirectlyfrom thechangein themixed-layerconcenCO2 via (10). For the early 1990s, s4.) Themagnitude of theanthropogenic carbon storage term trationof anthropogenic of anthropogenic CO2 wouldhave can be estimatedfrom the observedmeanpenetrationdepth the surfaceconcentration budget(Figure 7c). (MPD) [Broeckeret al., 1979]for anthropogenic CO2, which is defamed as: increased by 0.7 to 0.8/•molkg4 yr-• in ordertokeeppace withtheatmospheric pCO2increase of about1.2•atrnyr4. This givesan averagelocalstoragerateof 0.59 (+20%) mol m'2yr-•or240(+49) kmols-•whenintegrated overtheentire meo =fc, Atlantic Ocean surfacearea between10øS and 30øS (12.8 x 10•2m2). whereC,aandCzaretheconcentrations of anthropogenic CO2 This estimate, which is based on the framework of a vertical diffusion-advectionmodel [Gammon et al., 1982], in the mixed layer and at depthz, respectively.Using the ensemble of individualanthropogenic CO2profilesfrom each can be tested against the directly observed buildup of anthropogenic CO• betweentheGeochemical OceanSections Latitude(South) Latitude(South) 10 15 20 25 30 10 15 20 25 30 I I I I I I I I I I A.Contemporary (1990s) TCO 2 . Preindustrial TCO 2 - -1500 -1000 - -1500 - - -2000 - - -2500 -2500 - - -3000 -3000 - - -3500 -2000 o c- ' I I - - D.Dissolved Oxygen 2000 B.Anthropogenic TCO 2 0 1500 - -500 1000 - -1000 o c- o 500• - -1500 0 - I I I I I 10 15 20 25 30 Latitude(South) 10 15 20 25 c- -2000 30 Latitude(South) Figure 8. Mean transportversuslatitudeplots,with regression lines and 95% confidence intervals,basedon the results presented in Table4. Separate plotsarepresented for thefollowingparameters: (a) thecontemporary (1990s)inorganic carbon transport, (b) theanthropogenic carbontransport, (c) thepreindustrial inorganic carbontransport, and(d)thedissolved oxygen transport. Notethatthetransport axisforeachplotcovers a totalrange of2500lcmol s4 andthattheconvention usedissuch thata negativetransportdenotestransportsouthward. 494 HOLFORT ET AL.' SOUTH ATLANTIC CO2 TRANSPORT Study expedition oi•themid-1970s andtheWOCEexpeditionscally driven fluxes of O2 and CO2 have oppositesigns, of the 1990s. Wallace et al. [1996] conducted such an analysis for thewesternbasinof theSouthAtlantic(5ø- 35øS, west of 20øW) using a multivariate,time seriesanalysis technique to comparetheTCO2 datacollectedduringthetwo samplingperiods.(Notethatour datasuggest thatthismore localizedregionhasa meanpenetration depththatis indistinguishable fromthe 10ø- 30øSaverage.)On thebasisof this empiricalapproach,the inventorychangeover 20 yearsis estimated to be 11.1 mol m-2whichgivesa storage rateof 0.56molm-2, identical to theone-dimensional model-derived estimate derived above. whereasfluxesdrivenby thermalprocesses operatein the samedirection.In thisregard,our estimates of themeridional oxygentransport(Table 4 and Figure 8d) indicateconvergenceof the net southwardO2 tramport,whichwe assume from direct transportestimates.In addition,the possible effectsof seasonality on the convergence estimates require further examination. 7. PreviousCO2 Transport Estimates There have been severalearlier publishedattemptsto estimatethe meridionaltramportof inorganiccarbonin the 10øSand30øSof 330(_+380)kmolS'1in thecontemporaryAtlantic Ocean. These used several different data sets, oceanis thereforesimilarin magnitudeto thisanthropogenic assumptions, andcalculationprocedures andare summarized carbonstorageterm. Further, the difference (90 _+380 kmol in Table5. A veryimportant distinction between ourtransport s'l) between thecontemporary inorganic carbon convergenceestimatesand many found in the earlier literature is our andtheanthropogenic carbonstorageis reasonably consistent explicitincorporation of the net meridionalmasstramports with a completelyindependent estimateof the net annualsea- requiredto satisfysalt and freshwaterbalances.$toll et al. to-airfluxforthisregionof 264kmols-1derived fromsurface [1996]notedthe significance of the CO2 transport contribupCO2 data[Takahashiet al., 1997]. The uncertainty of the tionwhicharisesfromnetmeridional masstransport through latterestimatecouldbe ashighas75 %. Thiscorrespondence Bering Strait. They also noted that changesin the mass suggests thatanyadditionalnetsourcesandsinksof inorganic transportas a functionof latitude,resultingfrom freshwater carbonin thisregion,suchas the localnetproduction/respi- additionor removal, drive a significantmeridionalcarbon rationof organiccarbonor dissolution of calciumcarbonate, transportdivergence. arenotsignificantly different thanzero(174_+400kmols-l). FollowingLundbergand Haugan [1996] and $toll et al. Figure 8b and Table 4 showthat during the 1990sthere [1996],we haveexplicitlyincludedthisnetmasstransport in wasa significant(p = 0.05) divergence of theanthropogenic ourestimates of themeridionalCO2transport. ThiscontribuCO2 transportbetween10øSand30øSamounting to ~320 tion wasnot, however,includedin the CO2 transport esti(+ 128)kmols-1.Anair-to-sea fluxof anthropogenic CO2of matesgivenby Breweret al. [1989]andBroeckerandPeng 560 (_+137)kmol s'• wouldbe requiredto balancethe [1992]. It was alsonot incorporatedin transportestimates combined anthropogeniccarbon storage and transport reportedin preliminarymeetingabstracts by Holfort et al. divergenceterms,assuming no additionalsources or sinks.In [1994]andRobbins[1994]. The latterwascitedby Sarmiento otherwords,the regionalsea-to-airflux of CO2 duringthe et al. [1995] in their estimationof the inorganiccarbon 1990swas approximately 560 kmol s'1 smallerthanthe transportdivergencebetweenBering Strait and the South corresponding sea-to-airflux of the preindustrialera. Atlantic. Martel and Wunsch[1993] accountedfor runoff, Subtraction of theanthropogenic carbontransportfromthe evaporation,and precipitationwithin their model domain contemporary transportimpliesa statistically significant (p = (12ø-60øNin theAtlanticOcean)butapparently chosenotto 0.1) localconvergence of Tc between11øSand30øSfor the includethe salt and mass transportthroughtheir model preindustrialera (Table4 andFigure8c). Duringthisperiod, domainarisingfrom the BeringStraitflow andrunoffnorth we assumethatthe overallinorganiccarboninventorywasin of 60øN. KeelingandPeng[1995]accounted for theBering steadystate(zero storageterm), andif thebiologicalproduc- Straitcontribution to theNorthAtlanticphosphate balancebut tivity of the regionwas comparableto thatof the contempo- chosenotto includeitscontribution to thetotalmasstransport rary ocean,the convergence wouldhavebeenbalancedby a when calculatingthe inorganiccarbontramportacrossthe considerable netsea-to-air fluxof CO2.Theaverage preindus- equator.Carboncyclemodelsimulations [e.g., Sarmiento et trial inorganic carbonconvergence (32.6 _+17 kmols'1per al., 1995] do not alwaysresolvethiscontributionbecausethe degreeof latitude)impliesa totalconvergence of 650 (_+340) transportthroughthe BeringStraitis not alwaysresolved. kmol s-1 between10øS and 30øS. Alternatively,taking To someextent,thechoiceof whetherto includeor ignore Takahashi et al.'s [1997]estimate of 264 (_+198)kmols'1for this contributionto the net carbontransportis a matterof the contemporarysea-to-airflux of CO2 (see above)and philosophy andis contingent on the specificquestions being adding 560(_+137)kmols'1to compensate fortheair-to-sea askedof thetramportestimates.Oneof themajorreasons for transferof anthropogenic CO2 impliesa preindustrial sea-to- examiningcarbontransportdivergences is to obtainestimates air CO2 flux of 824 (_+240)kmols-1.Onceagain,thisis of the regionaldistributionof the net air-seaflux of carbon closely comparableto the preindustrialinorganiccarbon (andoxygen).With thisspecificgoalin mind,it isjustifiable convergence suggesting thatlocalnet production/respirationto ignorethe inorganiccarbontransportwhichis tied to the of organiccarbonor dissolutionof calciumcarbonatewas net salt transportacrosssections.This componentof the relativelysmallcomparedto thepreindustrial sea-to-airflux. inorganiccarbontransport,while large, can be viewed as Keelingand Peng [1995] haveshownthatthejoint inter- constant across all Atlantic sections and therefore does not pretationof meridionaltransportsof both02 and CO2 can affector reflectthe air-seaflux of CO2 or 02. However,if distinguish thephysicalversusbiologicalmechanisms respon- directcomparisons are madewith the net transportof inorsiblefor the fluxesof thesegases.As theypointout, biologi- ganiccarbonthroughtheBeringStrait,suchasweremadeby The very slightconvergenceof inorganiccarbonbetween HOLFORT ET AL.: SOUTH ATLANTIC CO2 TRANSPORT Sarmientaet al. [1995], then it is essentialthat the contribushouldnot have changedsince the preindustrialera. The observation thatthetransportof bothgasesin thepreindustrial ocean was convergentis qualitatively consistentwith a thermalforcingmechanism drivingtheeffluxof both02 and CO2acrossthe air-seainterfacein thisregion. Our heattransportcalculations (Table4) indicatesignificant (p < 0.1) divergenceof the northwardheat transport between 10øSand30øSaveraging 0.012(_+0.006)x l015W per degreeof latitude,implyingadditionof heatto the water colunto. Watsonet al. [1995] have estimatedthe maximum air-seaCO2 flux drivenby the heat transportof the North 495 precipitation, whichcansignificantly affectthedivergence of the net meridionalmasstransport[Stall et al., 1996] and which mustbe accountedfor when estimatingcarbontrans- portdivergences..As notedbyStalletal. [1996],thepioneeringanalysis byBreweretal. [1989]wasbasedonanassumption of zero net meridionalmasstransportand thereforedid not include this effect. Contraryto a statement by Stall et al. [1996], the CO2 transport estimates of BraeckerandPeng[1992]andKeeling and Peng [1995] do implicitlycorrectfor this effect by normalizing theobserved inorganic carbonconcentrations to constant salinity.Throughthisnormalization, theadditionof Atlanticoceanto be ~0.75Gt C yr'• or ~2000kmols4 of CO2uptakefor eachpetawatt (10isW) of northward heat variable levels of freshwaterto different water masses(and flux divergenceis harderto explainexceptas the resultof (1) uncertaintyin thevarioustransportandconvergence estimates themselves(note the large error bars for the convergence estimates and,particularly,thepossibleinfluenceof seasonality of thetransports,section4.3); (2) violationof theassumption of no storagefor dissolvedoxygen;or (3) an "extra" contribution to a predominantly thermallydrivenconvergence arisingfrom somebiologicalmechanism(e.g., a divergence of the organiccarbontransport). On balance, the qualitativepicture is consistentwith independent estimates of theair-seaCO2flux, andtheoverall budgetfor preindustrialcarbonsuggests a relativelyminor role for biologicallydriven convergence/divergence in this particularregion.However,it shouldbe clearfrom theabove that measurementsof the total organiccarbonconcentration of seawater,which were unfortunatelynot availablefor the sectionsused in this study, would provide independent comtraintson the air-seaCO2 exchangescenariosinferred tion of this componentof the transportbe includedin the transportestimatesat all latitudes. Of potentially more significanceis the effect of net freshwateradditionor removalvia runoffor evaporation and netmass flowthrough a section which is'required tobalance hencedilutionof inorganiccarbon)is compensated for. In transport.Applyingtheir scalingto our observeddivergence principle,thiscorrection shouldservethe samepurposeas of the northward heat transportbetween 10øS and 30øS keepingtrackof thefreshwater transport divergence; how(Table4) givesa maximumthermallydrivenCO2sea-to-air ever, in practice,the efficacyof this "normalization" apfluxof (2000x 20 x 0.012)or 480(_+240)kmols'l, which proachhasnotbeenindependently verified.Thatis, asit was is quiteconsistent with the observedpreindustrialinorganic applied,did the salinitynormalization reflecta reasonable carbonconvergence andestimated preindustrial sea-to-airflux freshwater transport distribution? In contrast,wehavechosen of CO2 (see above). A similar calculationfor dissolved to explicitlyconsider the salt,freshwater, andcarbontransoxygen givesa maximum sea-to-air 02fluxof ~1500kmols-• portsratherthancalculating thetransport of salinity-normalPW-•ofheattransport divergence, whichimplies a maximum ized quantities.Using a measurement-based salt transport constraint,estimatedbaroclinic, Ekman, and barotropic thermallydrivensea-to-air02 flux of (1500 x 20 x 0.012) or 360 (_+ 180) kmol s-• for the samelatituderange.The transports, andthe observed salinitydistribution, we were observedconvergence of oxygen,however,was1180(_+460) ableto directlyestimate thefreshwater transport divergence. kmol s-• between 10øS and 30 øS. This was consistentwith independentestimatesof this Hencethe observedconvergence of bothgasesis qualitaquantity[Baumgartner andReichel,1975](seesection 4.4) tively consistentwith the signof the observedheattransport therebyincreasing our confidence thatthe freshwater transdivergence.In the caseof CO2,themagnitudeof the converportdivergence is beingcorrectly represented whenestimatgenceis alsoquantitatively consistent with themaximumthat ingtheinorganiccarbontransport divergence. wouldbe predictedtheoreticallyfrom the observedheatflux Thisdiscussion emphasizes thewidevarietyof approaches divergence, whereas for dissolvedoxygen the observed and philosophies that havebeenpromulgated in orderto calculatecarbontransports. Not surprisingly, thisvarietyhas convergence is 2-3 timesthetheoretical prediction.Giventhat led to confusions of interpretation and a risk of comparing air-seaequilibrationof O2 is an order of magnitudemore unlike transportestimates. It is thereforeimportantto rapidthanthat of CO2, it is perhapsnot surprisingthat gas reiteratethat we have reportedthe total net transportof exchangewouldcreatea strongerconvergence signalfor 02 These than for CO2. However, an observedoxygenconvergence inorganiccarbonacrossthe SouthAtlanticsections. estimates incorporate thecarbontransport associated with a larger than that predictedon the basisof the observedheat the salt inputthroughBeringStrait. They thereforealso reflect the meridionallyvaryingmasstransportwhich is associatedwith freshwater addition and removal. Table5 highlights someotherwaysin whichourapproach to calculating inorganiccarbontransport differsfromprior estimates.Notably, our estimateis basedon a very large, high-quality dataset,includingfour sections of TCO2data which have been interpolatedonto the CTD data using multipleregression. Manyof theearlierestimates werebased on limiteddatasets,oftenrelyingon bulk averageconcentrationsfor differerawater massesor layers. Severalanalyses havereliedon bulk overturningcalculations in whichthenet massfluxin anupperlayeranda deeplayeraremultiplied by zonallyaveraged propertyconcentrations for therespective layers. The potentialrisksassociated with estimatesthat are basedon smoothed fieldsare discussed in detailby Rintoul and Wunsch [1991], who found that accurate estimatesof nutrienttransportscanrequirevery detailedestimatesof the circulationandpropertyfields. We examinedthe differencebetweenthe Tc calculated usingour full inversemodel solutionsand that derivedfrom 496 HOLFORTET AL.' SOUTHATLANTIC CO2TRANSPORT a zonallyaveragedestimateof thetransportfor theMeteor28 (11øS) and Meteor 22 (30øS) sections.For the zonally averagedTc, we calculatedthe sum, over all 19 vertical layers in the model, of the productof a layer-averagenet masstransportanda layer-average TCO2 concentration. At 30øS, the zonally averagedsolutionwas only 100-250 kmol andwas-2150_+200kmols'• across 20øS.Thistransport includesa large contribution associated with a barotropic velocityrequiredto satisfythe salt flux constraintmeasured at the Bering Strait. While estimationof this contributionis notessential for assessing theinorganiccarbonuptakeby the NorthAtlanticoverturning circulation andwasnotincorpos4 (10%) differentfromthefullmodelsolution. However,at ratedintoseveralearlierestimates of oceancarbontransport, 11øSthezonallyaveraged solution wasalmost1000kmols-• it is criticalfor assessing the oceancarbontransportdiveror 50% lower thanthe full solution.Inspectionof the solu- gencebetweenthe BeringStraitand the SouthAriantic.The tionssuggested thatthelargedifferenceat 11øSwastheresult principalsourcesof uncertainty in the transportestimates of a strongwesternboundarysubsurface currentin thedepth originatewith the Ekman component of the transport, range 50-800 m at that latitude [cf. Schottet al., 1995] particularlyat low latitudes,togetherwith uncertaintiesin coupled with a large difference in TCO2 concentration determining the barotropic component of thevelocityfield. between the western and eastern boundaries. Thebarotropic transport uncertainty wasreduced through use We also comparedthe Tc derivedfrom the full inverse of aninversemodelandcouldprobably bereduced furtherby modelsolutionswith that derivedfrom a two-layerzonally the use of more sophisticated inversecalculations and addiaveragedoverturningcalculationsimilar to that used by tional constraints.To lower the uncertaintyin the Ekman Broeckerand Peng [1992] andKeelingandPeng[1995]. We transport,a betterknowledgeof theannualwindstresswill be definedthe upperlayer as o2 <36.8 and definedthe lower required.A better understandingof the effectsof seasonal layer as o2 > 36.8. The magnitudeof the overturning variationsof the carbontransportis required. circulationimpliedby thesecalculations rangedfrom 11 to 14 By estimating theanthropogenic component of theTCO2 x 10• kg s-1,whichis closeto thatassumed byKeelingand distribution, it was shown that there was a net northward Peng[1995]butmuchlowerthanthe20x 10• kgs-•assumed transportof anthropogenic carbonacross20øS of 430 kmol by Broeckerand Peng [1992]. As with the zonallyaveraged s-•.Thistramport increased toward thenorth.Bydifference, solutiondiscussedabove, at 30øS the simpleoverturning the preindustrialsouthwardtransportof inorganiccarbon calculation gavetransports whichwereonly200-300kmols-• across 20øSwas-2580kmols-• . Thepreindustrial transports (10-15%) lower than the full model solution.However, at 11øSthe overturningcalculationgavea transportwhichwas shownetconvergence of bothCO2and02 between10øSand 30øS, suggesting that this region was a net sourceof both 700-900kmols-• or ~40%lowerthanthatderived usingthe gasesfor theatmosphere duringthepreindustrial era, consisfull model. Obviously, caution must be exercisedwhen tent with thermalforcingof the gas flux. Our transport attemptingto estimatethe transportof inorganiccarbon; estimates alsosuggest thattransportof anthropogenic carbon undercertaincircumstances it appearsthatsimpleaveraging by the upperlimb of the meridionaloverturningcirculation of the velocityand concentration fieldscanprovideaccurate may supplya significant fractionof the anthropogenic CO2 carbontransports. However, in other circumstances such simplifiedcalculations may be subjectto largeerrors. Our analysissharesmanysimilaritieswiththework of $toll et al. [1996], whoestimated the CO2transportacrossa single sectionbetweenGreenlandandIreland. Our analysisbenefits from having been able to samplethe boundarycurrents, whereasice precludedStoll et al. from samplingwithin the East Greenland Current. Our multiple zonal sectionsalso allow for a more detailedinverseanalysis.The SouthAtlantic sectionsbenefit from being closed, whereas the section occupiedby $toll et al. [1996] couldnot explicitlyaddress transportvia the CanadianArctic Archipelago,the English Channel, and the Irish Sea. For example, Lundbergand Haugan [1996] estimatethe net masstransportthroughthe Canadian Archipelago to be 1.4 x 10• kg s4 witha mean which is currently being storedwithin the North Atlantic Ocean. Appendix The constraints usedin the inversemodelanalysiswereas follows: 1. The total salttransportacrossthe SouthAtlanticzonal sections is26.7x 106kgs-•. Thisconstraint isbased onthe salttransportthroughtheBeringStraitandtheassumption of a steadystatesalt balancewithin the North Atlantic. Notethatthisisa departure fromthemorecommonly used constraint of net masstransport usedfor heattransport calculations. 2. Salt is conservedin boxesboundedby the different sections and/orthecontinents in thehorizontal planeand by specifieddensitysurfacesin the verticalplane.The verticallayersincludethoseusedby MacDonald[1993] TCO 2concentration of~2070 •tmol kg-•s._• This implies a nonnegligibletransportof ~3000 kmol via this route. Sections whichresolvethistransport pathway(e.g., sections acrossthe LabradorSea) shouldbe incorporated into any future inverseanalysisof net inorganiccarbontransport within the northern North Atlantic. andRoeromich[1983] and are listedin the captionto Figure 4. 3. Thereis no netmeridionalflow withinlayersdeeper than 3000 dbar in the eastern basin of the South Atlantic 8. Summary The transportof inorganiccarbonacrosssix zonalsections occupiedbetween11øSand30øSin theSouthAtlanticOcean hasbeenestimated. During the early 1990s,the transport wasdirectedsouthward,decreased slightlytowardthe south, becauseof the blockingof suchflowsby the Walvisand Mid-AtlanticRidges. 4. The transportof the Brazil currentat 30øS(vertical extent of0-600m)istakentobe10x 109kgs-1southward, andthe flow of AntarcticBottomWater (04 > 45.92) through theVemaChannel is5 x 109kgs-•northward. The HOLFORT ET AL.' SOUTH ATLANTIC CO2 TRANSPORT 497 values of these transportswere determinedfrom direct References current measurements[Holfort, 1994; Holfort et al., Anderson,L. G., D. Dyrssen,andE. P. Jones, An assessment of manuscriptin preparation,1998). the transportof atmosphericCO2 into the Arctic Ocean, J. 5. KeelingandPeng [1995] appliedtheconstraint thatthe Geophys.Res., 95, 1703-1711,1990. net meridionaltransportof phosphate acrossthe equator Baumgartner, A., andE. Reichel,Die Weltwasserbilanz, 197pp., Oldenburg,Munich, Germany,1975. shouldequalthephosphate transportthroughBeringStrait (about i.7 kmol s-:). Use of a similar constraintin our Bates, N. R., A. F. Michaels, and A. H.-Knap, Seasonaland interannual variabilityof oceaniccarbondioxidespeciesat the inverse calculationsignificantlyrestrictedthe range of U.S. JGOFSBermudaAtlanticTime-SeriesStudy(BATS) Site, valid solutionsand better constrainedthe heat, water and Deep-SeaRes., II, 43, 347-383, 1996. carbontransports,while only slightlyalteringtheir mean Boebel, O., C. Schmidt, and W. Zenk, Flow and recirculation of Antarctic Intermediate Water acrossthe Rio Grande Rise, J. values(Holfort et al., manuscriptin preparation,1998). Geophys.Res., !02, 20967-20986,1997. However, it is probablymore appropriateto assumea balancefor totalphosphorus in theAtlantic,whichsuggests Brewer, P. G., Direct observationof the oceanicCO2 increase, Geophys.Res.Lett., 5, 997-1000, 1978. that the transport of organic phosphorusmust also be Brewer,P. G., andJ. C. Goldman,Alkalinitychanges generated by considered.There are surprisinglyfew measuresof total phytoplankton growth,Limnol.Oceanogr.,21(1), 108-117,1976. phosphorus in theopenocean,yet dissolved organicforms Brewer, P. G., C. Goyet, and D. Dyrssen, Carbon dioxide transportby ocean currentsat 25øN latitude in the Atlantic of phosphorus constitute 50-100% of thephosphorus in the Ocean, Science, 246, 477-479, 1989. ocean'ssurfacelayer. Measurementsat the Hawaii Ocean P. G., D. M. Glover,C. Goyet,andD. K. Shafer,ThepH Time Series Site [e.g., Karl and Tien, 1997] show dis- Brewer, of the North AtlanticOcean:Improvements to the globalmodel solved organic phosphorusconcentrationsof 0.1 - 0.3 for soundadsorptionin seawater,J. Geophys. Res., 100, 8761- 8776, 1995. •mol kg-• withintheupper500 m, dropping to < 0.05 transportof •mol kg-• below1000m. If suchlevelsaretypicalof the Broecker,W. S., andT.-H. Peng, Interhemispheric SouthAtlantic,thereis a potentialnetnorthwardtransport carbondioxideby oceancirculation,Nature,356, 587-589, 1992. W. S., T. Takahashi,H. J. Simpson,and T.-H. Peng, of organicphosphorus of theorderof 10 x 109 kg S-• X Broecker, Fate of fossilfuel carbondioxideandthe globalcarbonbudget, 0.25/•molkg4 or 2.5 kmols'• associated withtransport Science, 206, 409, 1979. withinthe upperlayersof the ocean.This is presumably Bryan, K., Measurements of meridionalheat transportby ocean currents,J. Geophys.Res., 67, 3403-3414,1962. balancedby a net southward transportof inorganicphosphorus(phosphate)which is in additionto the phosphate Chen, C.-T. A., and F. J. Millero, Gradual increase of oceanic CO2, Nature,277, 205-206, 1979. transportedthroughthe BeringStrait.Withoutdataon the Coachman, L. K., andK. Aagard,Transports throughBeringStrait: concentrationof organicphosphorus,thereis considerable Annualandinterannualvariability,J. Geophys.Res., 93, 15535uncertaintyin how to exactlyspecifythe phosphatetrans15539, 1988. port constraint.Hence duringthe inverseanalysis,three Conway,T. J., P. P. Tans,andL. S. Waterman,Atmospheric CO2 recordsfrom sitesin theNOAA/CMDL air sampling network,in separatevariantsof thephosphate transportconstraintwere Trends'93:A Compendium of Data on GlobalChange,editedby appliedwhichbounda reasonablesetof possibilities:(1) T. A. Bodenet al., Rep. ORNL/CDIAC-65, pp. 41-119,Carbon The phosphatetransportwas specifiedto be identical DioxideInf. andAnal. Cent., Oak RidgeNat. Lab., Oak Ridge, acrossall of the SouthAtlanticsections,but its magnitude Tenn., 1994. wasnot specified.The phosphate transportwhichresulted Dickson,R., J. Lazier, J. Meincke,P. Rhines,andJ. Swift, Long- term coordinatedchangesin the convectiveactivityof the North Atlantic,Prog. Oceanogr.,38, 241-295, 1997. phosphateEnting, I. G., C. M. Trudinger,and R. J. Francey, A synthesis from applyingthisvariantof the constraintwasgenerally between 4 and8 kmols-• southward.(2) A transport of 2 kmols-• southward across all sections was specified(very similarto the constraintusedby Keeling andPeng[1995]). (3) A netphosphate transportof 4 kmol s'• southwardacrossall sectionswas assumedin order to balancethe likely net northward transportof organic phosphorus as well as the BeringStraitinput. For the resultspresentedin this paper, mostweight was assigned to the4 kmols'• constraint. inversion of the concentration and0•3Cof atmospheric CO2, Tellus, Ser. B, 47, 35-52, 1995. Fu, L.-L., Thegeneralcirculation andmeridional heattransport of thesubtropical SouthAtlanticdetermined by inversemethods, J. Phys. Oceanogr.,11, 1171-1193, 1981. Gammon,R.H., J. Cline, and D. Wisegarver,Chlorofluoromethanes in the northeast Pacific Ocean: Measured vertical distributions andapplication astransient tracersof upperocean mixing. J. Geophys.Res., 87, 9441-9454, 1982. Goyet,C., andD. Davis,Estimation of totalCO2concentration throughout the watercolumn,DeepSeaRes.,Part I, 44, 859877, 1997. Acknowledgments. Thisworkwassupported by theU.S. Depart- Goyet,C., D. Davis,E. T. Peltzer,andP. G. Brewer,Developmentof improvedspacesamplingstrategies for oceanchemical mentof Energy(Officeof Biological andEnvironmental Research) undercontractDE-AC02-98CH00016 andby the GermanMinistry properties: Totalcarbondioxideanddissolved nitrate,Geophys. Res. Lett., 22, 945-948, 1995. of Science andTechnology (BMBF). The support of MichaelRiches andT. F. Stocker,An improved of DOE in allowingus to completethisanalysisis acknowledged. Gruber, N., J. L. Sarmiento, methodfor detectinganthropogenic CO2in the oceans,Global The authorsbenefittedfromdiscussions withandinputfromCharlie Biogeochem.Cycles,10(4), 809-837, 1996. Flagg,AlfredPutzka,PaulRobbins,RalphKeeling,ErnieLewis, of JorgeSarmiento, ScottDoney,Rik Wanninkhof, andtheDOE CO2 Hall, M., and H. L. Bryden, Direct estimatesandmechanisms oceanheattransport, DeepSeaRes.,PartA, 29, 339-359,1982. SurveyScienceTeam.Assistance with datacollection from Rick Normalmonthlywindstress Wilke, Craig Neill, andthe scientists and crew of the Meteoris Hellermann,S., andM. Rosenstein, gratefully acknowledged. Weparticularly thankTaroTakahashi and DavidChipmanfor generously providing access to theirexcellent SAVE data set. This is a WOCE and JGOFS contribution. overtheworldoceanwitherrorestimates, J. Phys.Oceanogr., 13, 1093-1104, 1983. Holfort,J., Grossrfiumige zirkulation undmeridionale transporte im 498 HOLFORT ET AL.: SOUTH ATLANTIC CO2 TRANSPORT Sfidatlantik,Ber. Inst. Meereskd. ChristianAlbrechtsUniv. Kiel, 260, 1994. Holfort, J., K. M. Johnson,D. W. R. Wallace, and B. Schneider, Trans.AGU, 75(3), OceanSci. Meet. Suppl., 161, 1994. Robbins,P.E., andH. L. Bryden,Direct observations of advective Johnson,K. M., K. D. Wills, D. B. Butler, W. K. Johnson,and C. Roether, W., and A. nutrientandoxygenfluxesat 24o N in thePacificOcean,Deep Sea Res., Part I, 41, 143-168, 1994. Meridional transportof total dissolvedinorganiccarbonin the Roemmich, D., Estimation of meridional heat flux in the North SouthAtlanticOcean(abstract),Eos Trans. AGU, 75(3), Ocean Atlanticby inversemethods, J. Phys. Oceanogr.,10, 1972Sci. Meet. Suppl., 161, 1994. 1983, 1980. Isemer,H. J., andL. Hasse, TheBunkerClimateAtlasof theNorth D., Thebalanceof geostrophic andEkmantransports in Atlantic Ocean,Vol. 2, Air-SeaInteractions,252 pp., Springer Roemmich, thetropicalAtlanticOcean,J. Phys.Oceanogr.,13, 1534-1539, Verlag, New York, 1985. 1983. Johnson, K.M., and D. W. R. Wallace, The single-operator Roemmich, D., and C. Wunsch, Two transatlantic sections: multiparameter metabolicanalyzerfor totalcarbondioxidewith coulometric detection, DOE Res. Sumre., 19, Carbon Dioxide meridionalcirculationand heat flux in the subtropical North AtlanticOcean,Deep SeaRes., Part A, 32, 619-664, 1985. Inf. Anal. Cent., Oak RidgeNat.Lab., Oak Ridge, Tenn., 1992. S. Wong, Coulometrictotalcarbondioxideanalysisfor marine studies: Maximizing the performanceof an automatedgas extractionsystemand coulometricdetector,Mar. Chem., 44, 167-187, 1993. Johnson,K. M., D. W. R. Wallace, R. J. Wilke, and C. Goyet, Carbondioxide,hydrographic, andchemicaldataobtained during the R/V Meteor cruise in the South Atlantic Ocean (WOCE sectionA9, February-March1991),Publ. 4416, CarbonDioxide Inf. Anal. Cent., Oak RidgeNat. Lab., Oak Ridge, Tenn., 1995. Johnson,K. M., B. Schneider,L. Mintrop, andD. W. R. Wallace, Carbondioxide,hydrographic, andchemicaldataobtained during the R/V Meteor cruise22/5 in the SouthAtlantic Ocean(WOCE sectionA10, December 1992 - January 1993), Publ., Carbon Dioxide Inf. Anal. Cent., Oak Ridge Nat. Lab., Oak Ridge, Tenn., in press,1998. Josey,S. A., E. C. Kent, D. Oakley,and P. K. Taylor, A new globalair-seaheatandmomentum flux climatology, Int. WOCE Newslett., 24, 3-5, 1996. Karl, D. M., and G. Tien, Temporal variability in dissolved phosphorus concentrations inthesubtropical NorthPacificOcean, Mar. Chem., 56, 77-96, 1997. Keeling,C. D., R. B. Bacastow,A. F. Carter, S. C. Piper, T. P. Whorf, M. Heimann, W. G. Mook, andH. Roeloffzen,A threedimensionalmodel of atmosphericCO: transportbased on observedwinds, $er., 1, Analysis of observationaldata, in Aspectsof ClimateVariabilityin the Pacificand the Western Americas,Geophys. Monogr.vol. 55, editedby D.H. Peterson, pp. 165-236,AGU, Washington, D.C., 1989. Keeling,R. F., andT.-H. Peng,Transportof heat,CO2and02 by the Atlantic's thermohaline circulation, Philos. Trans. R. Soc. London, Ser. B, 348, 133-142, 1995. Putzka, Transient-tracer information on ventilationand transportof SouthAtlanticWaters, in The South Atlantic.'PresentandPast Circulation,editedby G. Wefer et al., pp. 45-62, Springer-Verlag,New York, 1996. Sabine,C. L., D. W. R. Wallace,andF. J. Millero,. Surveyof CO2in the oceansrevealscluesaboutglobalcarboncycle,Eos Trans. AGU, 78(5), 49, 54-55, 1997. Sarmiento,J. L., R. Murnane,and C. le Qu6r6, Air-seaCO2 transferand the carbonbudgetof the North Atlantic, Philos. Trans. R. Soc. London, Ser. B, 348, 211-219, 1995. Saunders,P.M. and B. A. King, Oceanicfluxeson the WOCE All section,J. Phys. Oceanogr.,25(9), 1942-1958,1995. Schott, F., L. Skamma, and J. Fischer, The warm water inflow into the westerntropicalAtlanticboundaryregion, spring1994, J. Geophys.Res., 100, 24745-24760, 1995. Siedler,G., and W. Zenk, WOCE Sfidatlantik1991, ReiseNr. 15, 30. Dezember 1990 - 23 Mfirz 1991, METEOR-Ber., 92-1, 126 pp., Univ. Hamburg,Hamburg,Germany,1992. Siedler, G., W. Balzer, T. J. Mfiller, R. Onken, M. Rhein, and W. Zenk, WOCE SouthAtlantic1992,CruiseNo. 22, 22 September 1992 - 31 January1993, METEOR-Ber., 93-5, 131 pp., Univ. Hamburg,Hamburg,Germany,1993. Siedler, G., T. J. Mueller, R. Onken, M. Arhan, H. Mercier, B. A. King, andP.M. Saunders,The zonalWOCE sectionsin the South Atlantic, in The SouthAtlantic.' Present and Past Circula- tion, editedby G. Wefer et al., pp. 83-104, Springer-Verlag, New York, 1996. Stoll, M. H. C., H. M. van Aken, H. J. W. deBaar, and C. J. deBoer, Meridionalcarbondioxidetransportin the northern North Atlantic, Mar. Chem., 55, 205-216, 1996. Takahashi,T., R. A. Feely, R. Weiss, R. Wanninkhof,D. W. Chipman,S.C. Sutherland,andT. T. Takahashi,Globalair-sea Lundberg,L., andP.M. Haugan,A NordicSeas- ArcticOcean fluxof CO2:An estimate basedonmeasurements of sea-air pCO2 difference,in NAS ColloquiumVolumeon CarbonDioxideand carbonbudgetfrom volumeflows and inorganiccarbondata, ClimateChange,editedby C.D. Keeling,pp. 8292-8299,Natl. GlobalBiogeochem. Cycles,10(3), 493-510, 1996. Acad. Sci., Washington,D.C., 1997. MacDonald,A.M., Propertyfluxesat 30øSandtheirimplications Tans, P. P., I. Y. Fung, and T. Takahashi,Observationalconfor the Pacific-Indianthroughflowandtheglobalheatbudget,J. straintson the globalatmospheric CO2 budget,Science,247, Geophys. Res.,98, 6851-6868,1993. MacDonald, A.M., and C. Wunsch, Oceanicestimatesof global oceanheattransport,in US WOCEImplementation Report,Rep. 1431-1438, 1990. Wallace, D. W. R., Monitoringglobaloceancarboninventories, report,54 pp., OceanObs.Syst.Dev. Panel,Tex. A&M Univ., 8, pp. 28-30, TexasA&M Univ., CollegeStation,Tex., July 1996. CollegeStation,Tex., 1995. Martel, F., and C. Wunsch, The North Atlantic Circulationin the Wallace, D. W. R., P. Beining, andA. Putzka, Carbontetrachloride andchlorofluorocarbons in the SouthAtlanticOcean,19øS, early 1980's- An estimatefrom inversionof a finite-difference J. Geophys.Res., 99, 7803-7819, 1994. model,J. Phys.Oceanogr.,23, 898-924, 1993. Neftel, A. E., E. Moor, H. Oeschger,andB. Stauffer,Evidence Wallace,D. W. R., K. M. Johnson,J. Holfort, andB. Schneider, Detectionof the changingCO2 inventoryin the ocean,in 1996 frompolaricecoresfor theincrease of atmospheric CO2in the U.S. WOCEReport,U.S. WOCEImplement. Rep.8, pp. 31-33, pasttwo centuries, Nature,315, 45-47, 1985. U.S. WorldOceanCirc. Exp. Office,Dept.of Oceanogr., Tex. Perry,G. D., P. B. Duffy, andN. L. Miller, An extended dataset A&M Univ., CollegeStation,Tex., 1996. of river discharges for validationof generalcirculationmodels, Watson,A. J., P. D. Nightingale,andD. J. Cooper,Modelling J. Geophys. Res., 101, 21339-21349,1996. atmosphere-ocean CO2transfer,Philos. Trans.R. Soc.London, Reid, J. L., On thetotalgeostrophic circulationof theSouthAtlantic Ser. B, 348, 125-132, 1995. Ocean:Flow patterns,tracersandtransports, Prog. Oceanogr., Wunsch,C., The North Atlanticgeneralcirculationwestof 50øW 23, 149-244, 1989. Rintoul, S. R., and C. Wunsch,Mass, heat, oxygenand nutrient determinedby inversemethods,Rev. Geophys.,16, 583-620, 1978. fluxesandbudgetsin the North AtlanticOcean,DeepSeaRes., Wunsch, C., D. Hu, and B. Grant, Mass, heat, salt and nutrient 38, suppl.1, S355-S377,1991. fluxesin theSouthPacificOcean,.J. Phys.Oceanogr.,13, 725Robbins,P. E., Direct observations of the meridionaltransportof 753. 1983. totalinorganiccarbonin the southAtlanticOcean(abstract), Eos HOLFORTET AL.' SOUTHATLANTICCO2TRANSPORT 499 Zenk, W. and T. J. Mfiller, WOCE studiesin the SouthAtlantic, Cruise No. 28, 29 March - 14 June 1994, MEIEOR-Ber., 95-1 193 pp., Univ. Hamburg,Hamburg,Germany, 1995. J. Holfort, BrookhavenNationalLaboratory,PO Box 5000, Dept. of Applied Science, Upton,NY 11973-5000 D.W.R. Wallace,Brookhaven NationalLaboratory, PO Box 5000Dept. of AppliedScience,Upton,NY 11973-5000 (e-mail wallace•bnl.gov) (ReceivedOctober14, 1997, revised,April 22, 1997; acceptedMay 1, 1998.)
© Copyright 2026 Paperzz